Quantum computers could soon take a bigger role in training AI—without needing impossibly large quantum memories—raising fresh questions for cryptography and blockchain security, according to a New Scientist report summarizing a new study from Caltech, Google Quantum AI, MIT and startup Oratomic.
Big-data bottleneck
One fundamental obstacle has been getting massive datasets—think terabytes to petabytes—into a quantum processor. To harness quantum effects like superposition, classical data must be encoded into quantum states, and preparing those states traditionally demanded large amounts of quantum memory. That’s been a practical barrier to applying quantum advantage to machine learning and other data-heavy tasks.
A smarter feeding strategy
The new study proposes a different approach: instead of preloading an entire dataset into quantum memory, quantum states can be prepared on the fly during processing. This “bit-by-bit” feeding drastically reduces the amount of quantum memory required to leverage quantum effects, the researchers say. If validated in hardware, the method would let quantum systems operate on big datasets while using far less quantum storage than previously thought necessary.
What this means for performance
The researchers argue this could let quantum machines outperform classical systems on specific data-processing tasks with far fewer logical qubits than once assumed. They estimate that a fault-tolerant quantum computer with roughly 300 logical qubits might surpass classical computers on certain problems, and that even a device with about 60 logical qubits could begin to show advantages in some AI data-processing workloads. Those logical qubits are error-corrected units—reliable building blocks that real, large-scale quantum computers still lack.
Why crypto and blockchain watchers should care
Beyond AI, the work underscores how rapidly the quantum–AI nexus is evolving. As Oratomic CTO Hsin-Yuan Huang put it, “Machine learning is really utilized everywhere…in a world where we can build this [quantum computing] architecture, I feel like it can be applied whenever there’s massive datasets available.” Oratomic CEO Dolev Bluvstein reminded observers that quantum progress can accelerate quickly: people have been skeptical in the past, but hardware advances have surprised expectations before. That matters for crypto: if quantum hardware matures faster than anticipated, the threat to conventional cryptographic schemes—and by extension some blockchain systems—becomes more urgent, reinforcing the need for post-quantum cryptography planning.
AI helping quantum hardware too
The relationship goes both ways. AI tools are already helping scientists model and analyze complex quantum systems, which speeds up hardware development and the search for practical quantum applications. As Adrián Pérez-Salinas, professor of computational physics at ETH Zurich, put it, “The quantum machine is a very powerful device, but you do need to first feed it…this study talks about feeding and how it’s enough to load [data] bit by bit, without overfeeding the beast.”
Bottom line
The paper doesn’t promise immediate real-world breakthroughs—fault-tolerant logical-qubit machines at the 60–300 scale do not yet exist—but it provides a plausible pathway to make quantum advantage over classical systems more attainable for certain AI workloads. For the crypto community, it’s a timely reminder: keep watching quantum progress and accelerate migration plans to quantum-resistant cryptography.
Read more AI-generated news on: undefined/news
